A Novel Approach Using Serious Game Data to Predict the WISC-V Processing Speed Index in Children With...

Area of Science:

• Neuroscience
• Machine Learning
• Developmental Psychology

Background:

• Processing Speed Index (PSI) from the Korean Wechsler Intelligence Scale for Children-Fifth Edition (K-WISC-V) is a key indicator of cognitive function in children with ADHD.
• Current limitations in testing frequency hinder short-term monitoring of PSI, crucial for timely intervention in ADHD.
• There is a need for accessible, objective methods to predict PSI scores for better ADHD management.

Purpose of the Study:

• To develop a machine learning model for predicting PSI scores in children with ADHD.
• To utilize behavioral data from serious games for objective and accessible cognitive function monitoring.

Main Methods:

• Recruited 68 children (ages 6-13) diagnosed with ADHD; 59 participants were included after data exclusion.
• Administered K-WISC-V for initial PSI assessment, followed by 25 minutes of engagement with serious game content.
• Trained machine learning models using game session data and evaluated predictive performance with RMSE, MAE, and MAPE, employing 4-fold cross-validation.

Main Results:

• Support Vector Regression (SVR) showed strong performance individually.
• An ensemble model integrating AdaBoost, Elastic Net, and SVR achieved the best predictive accuracy with the lowest RMSE (10.072), MAE (6.798), and MAPE (6.611%).
• The ensemble model demonstrated highest accuracy for PSI scores around the mean (100), indicating clinical reliability.

Conclusions:

• The developed machine learning model offers a potential objective and accessible tool for monitoring cognitive function in children with ADHD.
• This approach can complement traditional assessments, enabling continuous tracking of symptom changes and personalized treatment planning.
• Further validation in diverse populations and examination of long-term feasibility in clinical settings are recommended.